Hacker News: hanzeweiasa

New comment by hanzeweiasa in "Is This the Dawn of the Tokenpocalypse?"

hanzeweiasa — Mon, 08 Jun 2026 07:24:44 +0000

The token consumption concern is real but I think the framing misses a key trend: specialized models for specific domains.

In legal tech, we run domain-specific models for contract review that use 90% fewer tokens than general-purpose LLMs because they understand legal document structure natively. The token cost per document dropped from dollars to cents.

The real "tokenpocalypse" is for use cases that try to do everything with one general model. As the ecosystem matures toward specialized tools (similar to how we got specialized IDEs for different programming languages), token efficiency improves dramatically.

The analogy holds: general-purpose models are like Swiss Army knives — useful but inefficient. Domain-specific models are like proper tools — more expensive upfront but vastly more efficient for their domain.

New comment by hanzeweiasa in "Office-open-xml-viewer: Office XML document viewer that renders to HTML Canvas"

hanzeweiasa — Mon, 08 Jun 2026 02:11:35 +0000

Office XML is surprisingly complex under the hood. The format packages multiple XML streams, relationships, and content types into a ZIP — making debugging without specialized tooling painful.

Rendering to HTML Canvas is a pragmatic choice. We work with legal documents daily and the fidelity gap between native Office rendering and HTML-based viewers is one of those "last 10%" problems that takes 90% of the effort. Things like tracked changes formatting, table layout inheritance, and nested content controls rarely render correctly in lightweight viewers.

For document-heavy workflows (legal, compliance, procurement), having a viewer that preserves structural fidelity — especially revision marks and annotations — is table stakes. Most web-based solutions we tested lost formatting on documents with complex nested structures.

Interesting approach. Does the Canvas rendering handle tracked changes and inline comments? That is where most viewers break down.

New comment by hanzeweiasa in "Uber's $1,500/month AI limit is a useful signal for AI tool pricing"

hanzeweiasa — Wed, 03 Jun 2026 13:37:32 +0000

Good point about the unit of consumption shifting from prompts to agent loops. That makes pricing even trickier for vertical-specific AI tools.

We see this firsthand building AI Workdeck (open-source AI workspace for legal teams). A single due diligence review might chain 20+ agent calls: OCR -> text extraction -> clause classification -> risk scoring -> evidence chain assembly. The user sees one action, but the backend burns through significant inference.

The interesting thing about vertical tools is the pricing model can be fundamentally different. Horizontal tools charge per seat or per token. But in legal, the value is in the document, not the seat. A lawyer reviewing a 500-page M&A file gets way more value than one reviewing a 2-page NDA.

Self-hosting changes the calculus too. Our users run on their own infra, so the AI cost is whatever their GPU costs. That makes $1,500/month caps less relevant and throughput optimization more important.